Analysis of language queries in restricted discourse areas

This thesis presents a novel model for analyzing queries of the users of spoken language systems in restricted discourse areas. Such systems are able to interact with the user by means of speech and provide information in specific domains. The queries which may be posed are modeled by means of a formal grammar in the discourse area "currency conversion". This grammar allows the automatic analysis of spoken input. The results of many scientific disciplines, such as linguistics, computer science, and cognitive science contribute to the model. Special attention is paid to the integration of various knowledge types (semantic, syntactic, and pragmatic information) into a consistent model. This leads to a grammar life cycle, which is composed of a sequence of steps: grammar specification, grammar mining, grammar design, and iterative grammar evaluation. The development of the grammar is illustrated with numerous examples from the discourse area "currency conversion". Finally, our approach is compared with currently existing systems, and the integration into the design of user-friendly man-machine interfaces is discussed